hep-th/0012251

CTP-MIT-3064
PUPT-1972
MRI-P-001201

String Field Theory Around the Tachyon Vacuum

arXiv:hep-th/0012251 v1 27 Dec 2000

Leonardo Rastellia, Ashoke Senb and Barton Zwiebachc

a
Department of Physics
Princeton University, Princeton, NJ 08540
E-mail: rastelli@feynman.princeton.edu
b
Harish-Chandra Research Institute1
Chhatnag Road, Jhusi, Allahabad 211019, INDIA
E-mail: asen@thwgs.cern.ch, sen@mri.ernet.in
c
Center for Theoretical Physics
Massachussetts Institute of Technology,
Cambridge, MA 02139, USA
E-mail: zwiebach@mitlns.mit.edu

Abstract

Assuming that around the tachyon vacuum the kinetic term of cubic open string field
theory is made purely of ghost operators we are led to gauge invariant actions which
manifestly implement the absence of open string dynamics around this vacuum. We test
this proposal by showing the existence of lump solutions of arbitrary codimension in this
string field theory. The key ingredients in this analysis are certain assumptions about
the analyticity properties of tachyon Green’s functions. With the help of some further
assumptions about the properties of these Green’s functions, we also calculate the ratios
of tensions of lump solutions of different dimensions. The result is in perfect agreement
with the known answers for the ratios of tensions of D-branes of different dimensions.

1
Formerly Mehta Research Institute of Mathematics and Mathematical Physics

1
Contents
1 Introduction and Summary 2

2 String Field Theory Around the Tachyon Vacuum 6

2.1 Algebraic structure before and after condensation . . . . . . . . . . . . . . 7
2.2 An ansatz for the SFT after condensation . . . . . . . . . . . . . . . . . . 9

3 The Lump Solutions and Their Tensions 11

3.1 Analysis of the tachyon effective action . . . . . . . . . . . . . . . . . . . . 12
3.2 Setup for lump solutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
3.3 Equations for lump solutions . . . . . . . . . . . . . . . . . . . . . . . . . . 16
3.4 Solving the lump equations . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
3.5 Computation of ratios of lump tensions . . . . . . . . . . . . . . . . . . . . 22

4 Further studies and open questions 24

4.1 On the uniqueness of Q . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
4.2 On the use of level truncation . . . . . . . . . . . . . . . . . . . . . . . . . 26
4.3 On gauge fixing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
4.4 On superstring field theory and closed strings . . . . . . . . . . . . . . . . 28
4.5 Concluding remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30

1 Introduction and Summary

The 26-dimensional bosonic string theory contains D-p-branes for all integer p ≤ 26. Each
of these D-p-branes has a tachyonic mode. It has been conjectured [1, 2] that there is a
local minimum of the tachyon potential which describes the closed string vacuum without
any D-brane. At this minimum the negative contribution from the tachyon potential
exactly cancels the tension of the D-brane. Furthermore, since the brane disappears, this
vacuum cannot support conventional open string excitations. It has also been conjectured
that a codimension q lump solution on the D-p-brane represents a D-(p − q)-brane in the
same theory. Support for these conjectures comes from the analysis of the world-sheet
theory [1, 3, 4, 5, 6, 7, 8], cubic open string field theory [9, 10, 11, 12, 13, 14, 15, 16, 17, 19,
20, 21, 22, 23, 24], noncommutative limit of the effective field theory of the tachyon [25,
26, 27, 28, 29, 30], background independent open string field theory [31, 32, 33, 34], as
well as various simple models of tachyon condensation [35, 36, 37, 38, 39, 40].
One of the promising approaches to the study of this subject matter involves the
use of cubic open string field theory (SFT) [41]. Indeed, much work has already been

2
done with it using the level truncation scheme [9, 10, 11]. While complete analytic so-
lutions of the equations of motion of this string field theory are still missing, the solu-
tions found by combination of analytic and numerical work gave some of the strongest
evidence for the conjectures, as well as demonstrated that string field theory contains
non-perturbative physics. More recently, work in boundary string field theory (B-SFT)
[42, 43, 44, 45, 46, 47], − a version of string field theory incorporating a certain degree
of background independence, − confirmed the form of the tachyon potentials proposed
in [37, 38] and established conclusively the energetics aspect of the tachyon conjectures
[31, 32, 33]. Whereas in conventional SFT the tachyon string field condenses, namely
an infinite number of scalars acquire expectation values, in B-SFT only the tachyon field
acquires a vacuum expectation value. This is quite fortunate, as it appears to be very
difficult to formulate B-SFT in generality. Due to this problem in B-SFT, at present
cubic string field theory seems to be the only complete framework where we can analyze
concretely the fate of the open string field excitations around the vacuum where the D-
brane has disappeared. Indeed, it seems to provide a completely non-singular framework
to study the tachyon vacuum. Starting from a D-brane background one appears to reach
the tachyon vacuum with room to spare and without any indications of singularities in
field variables.
In principle the analysis of the tachyon vacuum is straightforward. One must: (i) find
the classical solution Φ0 representing the vacuum with no D-brane, (ii) expand the string
field action setting Φ = Φ0 + Ψ, where Ψ is the fluctuation field, and (iii) analyze the
spectrum of Ψ using the resulting kinetic term. Nevertheless, in practice this has not been
simple to carry out, first and foremost because there is no known closed form expression
for Φ0 yet. The accurate numerical approximations to Φ0 may allow, however, an analysis
of the spectrum around the vacuum [17, 18].
In this paper we shall analyze the problem from a different angle. Instead of trying
to construct the classical solution Φ0 , expand the SFT action around Φ0 , and attempt
field redefinitions to bring the kinetic term to a simple form, we shall make an inspired
guess about the form of the SFT action expanded around the tachyon vacuum. Then we
will try to check that this action satisfies the various consistency requirements. Since the
string field theory action is cubic in the string field, shifting the string field by a classical
solution does not change the cubic interaction term. Thus we only need to guess the
quadratic term of the shifted action. The requirement that the new action is obtained
from the original one by a shift of the string field by a classical solution of the equations
of motion puts constraints on the possible choices of the quadratic term, − these basically
correspond to the requirement that the shifted action also has an infinite parameter gauge
invariance like the original SFT action. Besides these constraints the action must satisfy

3
the following additional requirements:

1. The kinetic operator must have vanishing cohomology. This would imply absence
of physical open string states around the tachyon vacuum.

2. The action must have classical solutions representing the original D-brane configu-
ration, as well as lump solutions representing D-branes of all lower dimensions.

Indeed, our work began with the simple realization that the reparametrization ghost
zero mode c0 was an obvious replacement for the BRST operator that would lead to
vanishing cohomology, and thus no physical states.2 More striking, however, is the fact
that the Riemann surface description of the star product makes it evident that c0 is a
derivation. Thus, we noted that replacing the BRST operator QB by Q = c0 would lead
to a string field action
 
1 1 1
S(Ψ) ≡ − 2 h Ψ , Q Ψi + h Ψ , Ψ ∗ Ψi , Q = c0 , (1.1)
g0 2 3
that would have the expected (vanishing) spectrum and would be gauge invariant. It is
straightforward to see that more general operators lead to the same conclusions. In fact
∞
X
Q= an Cn , Cn ≡ cn + (−1)n c−n , (1.2)
n=0

with arbitrary constants an ’s would also lead to gauge invariant actions with no physical
states. As we will show, many of these actions are related by field redefinitions.
In fact, long time ago, Horowitz et.al. [49] constructed a class of formal solutions of
the purely cubic string field theory action [50, 51] which lead to actions of the above form,
with an extra regularity condition Q|Ii = 0, where |Ii is the (formal) identity of the star
algebra. In particular, these authors noticed that such string field theory would describe
no physical excitations around the vacuum in question. Furthermore, since a subclass of
the above actions is obtained by shifting the string field in the purely cubic action, which
in turn is obtained by shifting the string field in the original SFT action, we can, at least
formally, regard the above actions as the result of shifting the string field in the original
SFT action.3
We will not be able to determine here the precise form and normalization of Q and
we take the above class of actions as confimation that the conditions of zero cohomology,
2
This possibility may have occurred to many physicists, but we heard it first from E. Witten [48].
3
These solutions, however, have the unwanted feature that at least formally they have the same energy
density as the original D-brane configuration. This is not what is expected of the solution describing the
tachyon vacuum. This would suggest that we only consider those Q’s which do not satisfy the regularity
condition Q|Ii = 0. In fact, c0 |Ii =
6 0.

4
gauge invariance and universality of the kinetic operator can be satisfied. There may even
be more general forms of kinetic operators built purely from the ghost sector and satisfying
these conditions. Such kinetic terms make the universality of the action expanded around
the tachyon vacuum manifest. Indeed, since the tachyon vacuum, representing the closed
string vacuum without any D-brane, should not depend on which D-brane configuration
we start from, we should expect that the string field theory action expanded around
this vacuum should lose all knowledge about the matter conformal field theory (CFT)
describing the original D-brane configuration.4
Given the class of actions exemplified by eqns. (1.1) and (1.2) the non-trivial part of
the problem is to construct the classical solutions in this string field theory representing
the original D-brane as well as various lump solutions representing the lower dimensional
D-branes, and compare their tensions with the known tensions of the D-branes. At present
we have only been able to partially determine the space-time dependence of the solutions
and to calculate the ratio of the tension of a p-brane lump to that of a (p + 1)-brane lump
for arbitrary integer p. Under certain assumptions which will be explained in section 3,
this computation can be done without a detailed knowledge of the quadratic term in the
action as long as it does not involve any matter sector operator. The final result for the
ratio of tensions involve certain function B(α) (defined in eqs.(3.34), (3.37), (3.40)) which,
if zero, will produce the desired value. Although we cannot at present give an analytic
proof of the vanishing of this function, we have checked numerically that this function
indeed vanishes to very high degree of accuracy. This in turn is evidence that our guess
about the SFT action expanded around the tachyon vacuum is the correct one.
The strategy that we follow for this computation is as follows. Since the kinetic term
and hence the propagator does not involve any matter operator we are able to find the
exact momentum dependence of an arbitrary n-tachyon Green’s function. From the gen-
erating functional W [J] of the tachyon Green’s functions we can find the tachyon effective
action by Legendre transformation. Missing the overall (n-dependent) normalization of
the n-tachyon Green’s function, we only have a partial knowledge of the effective action.
We now assume that the tachyon equations of motion derived from this effective action
have a space-time independent classical solution which can be identified with the original
D-brane. Using this assumption, together with certain other assumptions about the an-
alyticity of the generating functional of the tachyon Green’s function stated in section 3,
we show that the tachyon equation of motion also has lump solution of every codimen-
4
Since the interaction term involves correlation function in the combined matter-ghost CFT, it does
depend on the matter CFT. For this CFT we use the canonical choice corresponding to the maximal
dimension space-filling D-brane. However, since the three string vertex involves complete overlap of the
three strings the interaction can be made formally background independent by a rescaling of the string
fields.

5
sion. Some further assumptions allow us to calculate the ratios of tensions of these lumps
without knowing the detailed form of the effective potential. As already stated, the result
is in perfect agreement with the known answers for the ratios of tensions of D-branes.
Clearly many things remain to be done. First and foremost is to reproduce correctly
the overall normalization of the D-brane tension. For this we need to fix unambiguously
the form of the kinetic term Q in the shifted vacuum.5 Some aspects of our analyticity
assumptions may require more information about Q. The action around the tachyon
vacuum is expected to possess many novel properties. In particular many of the sym-
metries which are spontaneously broken in the background of a D-brane are expected to
be restored in this vacuum [52, 53, 54]. It will be interesting to analyse our proposal
to check these properties. Another important problem is to generalize this analysis to
the case of superstring field theory (SSFT) [55, 56, 57] where analogous conjectures ex-
ist [58, 59, 60, 61, 62, 63] and some have already been tested using the level truncation
scheme [64, 65, 66, 67]. It is again natural to conjecture that at the tachyon vacuum the
kinetic term does not involve any matter sector operator.
The paper is organised as follows. In section 2 we discuss the properties that we expect
of the open string field theory action expanded around the tachyon vacuum, and construct
examples of actions satisfying these properties. In section 3 we show how, starting from
this action, and assuming the existence of a space-time independent solution of the equa-
tions of motion describing the original D-brane, we can calculate the ratio of tensions of
the lump solutions of different dimensions. In the concluding section 4 we discuss various
issues which are important for further extension of our analysis. These include a study
of the uniqueness of the proposed action, the possibility of using level truncation scheme
for calculating the overall normalization of the lump tension, an analysis of gauge fixing,
generalization to superstrings, and a discussion on closed string states.

2 String Field Theory Around the Tachyon Vacuum

Here we begin by considering open string field theory formulated on the background of
a specific D-brane, focusing on its algebraic structure both before and after tachyon con-
densation to the vacuum where the brane disappears. After a discussion of the properties
5
While Q = 0, leading to the purely cubic action [50], is universal, the resulting background is not
suitable for the present analysis since with vanishing kinetic term the Feynman rules used for computing
the tachyon effective action are not well defined. In addition, as opposed to the string field tachyon
condensate, the string field taking the original D-brane configuration to the one with vanishing kinetic
term is annihilated by the BRST operator. Thus the purely cubic action is unlikely to represent the
tachyon vacuum.

6
expected from the kinetic term of the string field theory around the tachyon vacuum we
propose an ansatz consistent with such expectations.

2.1 Algebraic structure before and after condensation

We begin with the cubic string field theory action describing the world-volume theory of
a D-(N − 1)-brane:
 
1 1 1
S(Φ) = − h Φ , QB Φi + h Φ , Φ ∗ Φi , (2.1)
go2 2 3
where go is the open string coupling constant and Φ is the open string field, conventionally
taken to be Grassmann odd and of ghost number one for the classical action. In addition,
QB is the BRST operator, h·, ·i is a bilinear inner product based on BPZ conjugation and
∗ denotes star-multiplication of string fields. The consistency of this classical action is
guaranteed by the following identities involving the BRST operator

Q2B = 0,
QB (A ∗ B) = (QB A) ∗ B + (−1)A A ∗ (QB B) , (2.2)
h QB A, B i = −(−)A hA, QB Bi ,

and identities involving the inner product and the star operation

hA, Bi = (−)AB hB, Ai

h A , B ∗ C i = hA ∗ B , C i (2.3)
A ∗ (B ∗ C) = (A ∗ B) ∗ C .

In the sign factors, the exponents A, B, · · · denote the Grassmanality of the state, and
should be read as (−)A ≡ (−)ǫ(A) where ǫ(A) = 0 (mod 2) for A Grassmann even, and
ǫ(A) = 1 (mod 2) for A Grassmann odd. We also have:

ǫ(A ∗ B) = ǫ(A) + ǫ(B)

gh(A ∗ B) = gh(A) + gh(B) , (2.4)

where gh denotes ghost number, and we take the ghost number of the SL(2,R) vacuum
to be zero. Equations (2.2) and (2.3) guarantee that the above action is invariant under
the gauge transformations:

δΦ = QB Λ + Φ ∗ Λ − Λ ∗ Φ , (2.5)

for any Grassmann-even ghost-number zero state Λ.

7
 Let Φ0 be the string field configuration describing the tachyon vacuum, a solution of
the classical field equations following from the action in (2.1):

QB Φ0 + Φ0 ∗ Φ0 = 0 . (2.6)

e = Φ − Φ denotes the shifted open string field, then the cubic string field theory
If Φ 0
action expanded around the tachyon vacuum has the form:
 
e 1 1 e e + 1h Φ
e,Φ
e ∗ Φi
e .
S(Φ0 + Φ) = S(Φ0 ) − 2 h Φ , Q Φi (2.7)
go 2 3

Here S(Φ0 ) is a constant, which according to the energetics part of the tachyon conjectures
equals the mass M of the D-brane when the D-brane extends over a space-time of finite
volume.6 Indeed, the potential energy V (Φ0 ) = −S(Φ0 ) associated to this string field
configuration should equal minus the mass of the brane. The kinetic operator Q is given
in terms of QB and Φ0 as:

e =Q Φ
QΦ e +Φ ∗Φ
e +Φ
e ∗Φ . (2.8)
B 0 0

More generally, on arbitrary string fields one would define

QA = QB A + Φ0 ∗ A − (−1)A A ∗ Φ0 . (2.9)

The consistency of the action (2.7) is guaranteed from the consistency of the one in (2.1).
Since neither the inner product nor the star multiplication have changed, the identities in
(2.3) still hold. One can readily check that the identities in (2.2) hold when QB is replaced
by Q [41]. Just as (2.1) is invariant under the gauge trasformations (2.5), the action in
(2.7) is invariant under δ Φe = QΛ + Φe ∗Λ−Λ∗Φ e for any Grassmann-even ghost-number
zero state Λ.
Since the energy density of the brane represents a positive cosmological constant, it
is natural to add the constant −M = −S(Φ0 ) to (2.1). This will cancel the S(Φ0 ) term
in (2.7), and will make manifest the expected zero energy density in the final vacuum
without D-brane. For the analysis around this final vacuum it suffices therefore to study
the action  
e 1 1 e e 1 e e e
S0 (Φ) ≡ − 2 h Φ , Q Φi + h Φ , Φ ∗ Φi . (2.10)
go 2 3
6
As suggested in [10], we shall assume that the time interval has unit length so that the action can be
identified with the negative of the potential energy for static configurations.

8
2.2 An ansatz for the SFT after condensation
If we had a closed form solution Φ0 available, the problem of formulating SFT around the
tachyon vacuum would be significantly simplified, as we would only have to understand the
properties of the new kinetic operator Q in (2.9). In particular we would like to confirm
that its cohomology vanishes in accordance with the expectation that all conventional open
string excitations disappear in the tachyon vacuum. Even the numerical approximations
for Φ0 may give some indication if this is the case [9, 17, 18].
Previous experience with background deformations (small and large) in SFT indicates
that even if we knew Φ0 explicitly and constructed S0 (Φ) e using eq.(2.10), this may not
be the most convenient form of the action. Typically a nontrivial field redefinition is
necessary to bring the shifted SFT action to the canonical form representing the new
background [73]. In fact, in some cases, such as in the formulation of open SFT for D-
branes with various values of magnetic fields, it is simple to formulate the various SFT’s
directly [74], but the nontrivial classical solution relating theories with different magnetic
fields are not known. This suggests that if a simple form exists for the SFT action around
the tachyon vacuum it might be easier to guess it than to derive it.
In proposing a simple form of the tachyon action, we have in mind field redefinitions
of the action in (2.10) that leave the cubic term invariant but simplify the operator Q in
(2.9) by transforming it into a simpler operator Q. To this end we consider homogeneous
field redefinitions of the type
e = eK Ψ ,
Φ (2.11)
where K is a ghost number zero Grassmann even operator. In addition, we require
K(A ∗ B) = (KA) ∗ B + A ∗ (KB) ,
h KA, B i = −hA, KBi . (2.12)
These properties guarantee that the form of the cubic term is unchanged and that after
the field redefinition the action takes the form
 
1 1 1
S(Ψ) ≡ − 2 h Ψ , Q Ψi + h Ψ , Ψ ∗ Ψi , (2.13)
g0 2 3
where
Q = e−K QeK . (2.14)
Again, gauge invariance only requires:
Q2 = 0,
Q(A ∗ B) = (QA) ∗ B + (−1)A A ∗ (QB) , (2.15)
h QA, B i = −(−)A hA, QBi .

9
These identities hold by virtue of (2.12) and (2.14). We will proceed here postulating a
Q that satisfies these identities as well as other conditions, since lacking knowledge of Φ0
the above field redefinitions cannot be attempted.

The choice of Q will be required to satisfy the following properties:

• The operator Q must be of ghost number one and must satisfy the conditions (2.15)
that guarantee gauge invariance of the string action.

• The operator Q must have vanishing cohomology.

• The operator Q must be universal, namely, it must be possible to write without

reference to the brane boundary conformal field theory.

The simplest possibility would be to set Q = 0. This would result in the purely cubic
version of open string field theory [50]. Indeed, it has long been tempting to identify the
tachyon vacuum with a theory where the kinetic operator vanishes, especially because
lacking the kinetic term the string field gauge symmetries are not spontaneously broken.
Nevertheless, there are well-known complications with this identification. At least in
the construction of [50] the string field shift Φ relating the cubic to the purely cubic
SFT’s involves the subtle identity operator I and appears to satisfy QΦ = 0 as well as
Φ ∗ Φ = 0. The tachyon condensate definitely does not satisfy these two identities. We
therefore search for nonzero Q.
We can satisfy the three requirements by letting Q be constructed purely from ghost
operators. In particular we claim that the ghost number one operators

Cn ≡ cn + (−)n c−n , n = 0, 1, 2, · · · (2.16)

satisfy the properties

Cn Cn = 0,
Cn (A ∗ B) = (Cn A) ∗ B + (−1)A A ∗ (Cn B) , (2.17)
h Cn A, B i = −(−)A hA, Cn Bi .

The first property is manifest. The last property follows because under BPZ conjugation
cn → (−)n+1 c−n . The second property follows from the conservation law

hV3 |(Cn(1) + Cn(2) + Cn(3) ) = 0 , (2.18)

on the three string vertex [15]. These conservation laws arise from consideration of inte-
R
grals of the form dzc(z)ϕ(z) where ϕ(z)(dz)2 is a globally defined quadratic differential.

10
In fact, the conservation law for C0 arises from a familiar Jenkins-Strebel quadratic differ-
ential with a second order pole at each of the three puntures. This quadratic differential,
in fact, defines the geometry of the vertex. Its horizontal trajectories show the open
strings and their interaction.
Each of the operators Cn has vanishing cohomology since for each n the operator
Bn = 12 (bn + (−)n b−n ) satisfies {Cn , Bn } = 1. It then follows that whenever Cn ψ = 0, we
have ψ = {Cn , Bn }ψ = Cn (Bn ψ), showing that ψ is Cn trivial. Finally, since they are built
from ghost oscillators, all Cn ’s are manifestly universal.
It is clear from the structure of the conditions (2.15) that they are satisfied for the
general choice:
∞
X
Q= an Cn , (2.19)
n=0

where the an ’s are constant coefficients. We will discuss in section 4.1 how different
choices for these constants result in kinetic operators that are sometimes related by field
redefinitions.
A subset of the above kinetic operators were considered by Horowitz et.al.[49], as they
discussed how the purely cubic SFT could have solutions leading to forms different from
the conventional cubic SFT. In particular they noticed that operators built just with ghost
oscillators could appear as the kinetic term of an action obtained by shifting by a (formal)
solution of the cubic theory.
There may be other choices of Q satisfying all the requirements stated above. For-
tunately, our analysis of section 3 will not require the knowledge of the detailed form of
Q, as long as it does not involve any matter operators. To this end, it will be useful to
note that since Q does not involve matter operators, we can fix the gauge by choosing a
gauge fixing condition that also does not involve any matter operator. In such a gauge,
the propagator will factor into a non-trivial operator in the ghost sector, and the identity
operator in the matter sector. Some details of gauge fixing are discussed in section (4.3).

3 The Lump Solutions and Their Tensions

In this section we shall analyze classical solutions of the string field action (2.13) intro-
duced in the previous section. The analysis will proceed in three steps: 1) analysis of the
tachyon effective action, 2) construction of the lump solution, and 3) computation of the
ratio of the tensions of the lump solutions.

11
3.1 Analysis of the tachyon effective action
We shall consider a situation where we start with an unstable D-(N − 1) brane of the
bosonic string theory, and consider the string field theory action expanded around the
tachyon vacuum of this theory. According to our postulates of the previous section this
action will have the form given in eq.(2.13). Given this action and a suitable gauge fixing
which does not involve any matter operator, the propagator involves purely ghost sector
operators. Since both the vertex and the propagator now factorize into matter sector
contribution and ghost sector contribution, a general N-point function will also factorize
into a matter sector contribution and a ghost sector contribution.
We shall now concentrate on the n-tachyon Green’s function. If we denote by J(p)
the current coupling to a tachyon of momentum (−p), then the generating functional of
the tachyon Green’s functions has the form:
∞
X1 Z N N (n)
Xn
W [J] = d p1 . . . d pn g (p1 , . . . pn )J(p1 ) . . . J(pn )δ( pi ) . (3.1)
n=2 n! i=1
P
Here g (n) (p1 , . . . pn )δ( ni=1 pi ) is the n-tachyon off-shell amplitude obtained by adding all
tree level connected Feynman diagrams. Note that since we consider a propagator that
involves just simple ghost factor, its world sheet interpretation is that of gluing with a zero
length strip and inserting a ghost operator. Thus the propagator collapses to an overlap
with a ghost insertion.7 Recall that the cubic vertex corresponds to the symmetric gluing
of three semi-infinite strips across the open edges, half edges at a time. It follows that for
n external tachyons any connected Feynman graph constructed with this vertex and the
overlap propagator will simply be built by the symmetric gluing of n-semiinfinite strips
across their open edges, half edges at a time. Mapping this world sheet into a unit disk,
each strip turns into a wedge of angle 2π/n. Except for the ghost insertions this can be
viewed as the iterated star product of n tachyon vertex operators. In fact, such iterated
products are precisely the ones used in the WZW-like superstring field theory [55] as
discussed in detail in [65].
It follows from the above remarks that the n-tachyon off shell amplitude can be written
as the following correlation function on a unit disk:
n
X n
Y (n)
g (n) (p1 , . . . pn )δ( pi ) = C n h (fk ◦ eipk ·X(0) )i , (3.2)
i=1 k=1

where h· · ·i denotes matter sector correlation function, Cn is a constant representing con-

(n)
tribution from the ghost sector correlator, fk ◦ eipk ·X(0) denotes the conformal transform
7
If the ghost kinetic operator does not have a simple inverse the fact remains that in the matter sector
the propagator will act as an overlap.

12
 (n)
of eipk ·X(0) by the function fk (z):
2 (n)
(n) (n)′
fk ◦ eipk ·X(0) = |fk (0)|pk eipk ·X(fk (0))
, (3.3)

and [65]:
 1 + iz 2/n
(n)
fk (z) = e2πi(k−1)/n . (3.4)
1 − iz
Using the relations:
(n) (n)′ 4
fk (0) = e2πi(k−1)/n , fk (0) = e2πi(k−1)/n i , (3.5)
n
and n n
Y (n) X Y (n) (n)
ipk ·X(fk (0))
h e i = δ( pk ) |fk (0) − fl (0)|pk ·pl , (3.6)
k=1 k=1 k6=l

we get
h n n
4 X X  π i
g (n) (p1 , . . . pn ) = Cn exp ln p2k + pk · pl ln 2 sin ( |k − l| ) , for n ≥ 3 .
n k=1 k,l=1 n
k6=l

(3.7)
(n)
While g given above is cyclically symmetric in p1 , . . . pn , only the fully symmetric part of
g (n) contributes to W . The tachyon propagator g (2) is a momentum independent constant.
We now perform a series of manipulations familiar in path integral quantization of field
theories. The tachyon effective action can be obtained from the generating functional W
by taking the Legendre transform. As usual, we introduce the classical field expectation
value as
φ(q) = φc [q, J] , (3.8)
where
X∞ Z Xn
δW 1
φc [−p1 , J] ≡ = dN p2 . . . dN pn g (n) (p1 , . . . pn )δ( pi )J(p2 ) . . . J(pn ) .
δJ(p1 ) n=2 (n − 1)! i=1
(3.9)
Note that φc [q, J] is a function of the momentum variable q and a functional of J(p).
In deriving the above equation we have used the cyclic symmetry of g (n) . The effective
action, a functional of classical fields φ(q), is given by,
Z
Γ[φ] = dN pJ(p)φ(−p) − W [J] , (3.10)

where J above is the source that gives rise to φ(q) through (3.8), (3.9). Since g (2) is
momentum independent, the coefficient of the quadratic term in Γ[φ] will be momen-
tum independent and the higher order terms will have exponential dependence on the

13
momenta. This resembles the action of the p-adic string theory expanded around the
tachyon vacuum[68, 35] after a momentum dependent rescaling of the tachyon field. As
in p-adic string theory we shall see that gaussians play an important role in constructing
lump solutions.
It follows from equation (3.10) that

δΓ[φ]
J(−p) = . (3.11)
δφ(p)

Thus, a solution to the equations of motion arising from setting the variation of the
effective action to zero amounts to setting the current J to zero. Therefore, classical
solutions are of the form:
φ(q) = φc [q, J(p) = 0]. (3.12)
Since φc [q, J] has a Taylor series expension in J starting with linear power, this would
seem to always give the trivial solution φ(q) = 0. However, W and φc have branch
points in the complex [J(p)] space, and by going around these branch points, one may
get φc [q, J(p) = 0] 6= 0. Every such path in the J space will give rise to a valid solution
of the equations of motion.8

3.2 Setup for lump solutions

We shall begin by examining translationally invariant solutions of the equations of motion.
(The original D-(N − 1)-brane configuration will be represented by such a solution.) In
momentum space a translationally invariant solution φ(p) is proportional to δ(p) where p
is an N-dimensional momentum vector. Consider eq. (3.9) with a delta function source
J(p) = uδ(p). We find
φc [q, uδ(p)] ≡ F (u)δ(q) , (3.13)
with ∞ ∞
X 1 X 1
n−1 (n)
F (u) = u g (0, . . . 0) = Cn un−1 , (3.14)
n=2 (n − 1)! n=2 (n − 1)!

where use was made of (3.7) for zero momentum. Now suppose that the function F (u)
has a branch point ub in the complex u plane such that as u returns to 0 after going
8
The situation can be illustrated by considering the zero momentum sector of the φ3 field theory. In
this case, at the tree level, Γ[φ] = − 21 m2 φ2 + 31 λφ3 . This gives J = −m2 φ + λφ2 , and φ = 2λ
1
(m2 ±
√
4
m + 4λJ) ≡ φc [J]. If we choose the − sign in front of the square root, φc [J] vanishes at J = 0
and has a Taylor series expansion in J starting at linear order. But going around the branch point at
J = −m4 /(4λ) in the complex J plane, we can move to the other branch with + sign in front of the
square root. Now if we return to the origin J = 0, we get φc [0] = m2 /λ. This is indeed the location of
an extremum of Γ[φc ] and hence represents a solution of the equations of motion.

14
around this branch point, F (u) returns to φ0 . Thus in this branch

lim F (u) = φ0 , (3.15)

u→0′

where 0′ is used to denote the origin on this branch. Hence limu→0′ φc (q, uδ(p)) = φ0 δ(q).
Eq.(3.12) then implies the existence of a translationally invariant solution of the equations
of motion of the form
φ(p) = φ0 δ(p) . (3.16)
Conversely, existence of a solution of the form given in (3.16) implies that the function
F (u) has a branch point ub in the complex u plane such that as u returns to 0 after going
around the branch point, F (u) returns to φ0 . We shall choose φ0 to be the value that
describes the original D-brane solution before tachyon condensation; such a solution exists
by assumption. In what follows, we shall assume that the branch point ub is the closest
singularity to the origin in the complex u plane, so that the growth of the coefficients
in the Taylor series expansion of F (u) is controlled by ub . In this case, the radius of
convergence of the Taylor series expansion of F (u) given in eq.(3.14) is |ub |.
Now suppose we find a pair of functions ψ(p) and χ(p) such that,

H(u, q) ≡ φc [q, uψ(p)] − F (u)χ(q) (3.17)

has the property that the rate of growth of the coefficients in the Taylor series expansion
of H(u, q) in u is slower than that of F (u)χ(q). In this case we can conclude that

1. H(u, q) does not have a singularity for |u| < |ub|.

2. H(u, q) either has no singularity at u = ub or its singularities at u = ub are non-

leading compared to that of F (u)χ(q).

We can now consider tracing the path around ub in the complex u-plane so that F (u)
approaches φ0 as u returns to the origin. If H(u, q) has no singularity at u = ub, then it
will return to zero. Otherwise it will return to some function h(q)

lim H(u, q) = h(q) . (3.18)

u→0′

As a result, on this branch (3.17) gives:

lim φc [q, uψ(p)] = φ0 χ(q) + h(q) . (3.19)

u→0′

Eq.(3.12) then implies that

φ(q) = φ0 χ(q) + h(q) , (3.20)

15
is a solution of the equations of motion. Thus given two functions ψ(q) and χ(q) satisfying
the property stated below eq.(3.17), and a translationally invariant solution φ0 δ(q), we
can find a space-time dependent solution φ0 χ(q) + h(q). Note that we have defined χ(q)
to be such that F (u)χ(q) is the leading contribution to φc (q, uψ(p)); so by definition, the
left-over piece H is sub-leading i.e. its power series expansion should converge faster than
that of F (u)χ(q). However, the existence of such a χ(q) is not guaranteed a priori since
the u-dependence of φc (q, uψ(p)) could be very different from that of φc (q, uδ(p)).
We shall now make the convergence condition precise. Let us define
R P
dN p2 . . . dN pn g (n) (p1 , . . . pn )δ( ni=1 pi )ψ(p2 ) . . . ψ(pn )
Rn (−p1 ) ≡ . (3.21)
g (n) (0, . . . 0)χ(−p1 )
Making use of this definition, and of eqs. (3.9), (3.14) and (3.21), equation (3.17) gives
∞
X un−1
H(u, −p) = χ(−p) Cn [ Rn (−p) − 1 ] . (3.22)
n=2 (n − 1)!

We need to ensure that the Taylor series expansion of H(u, q) converges more rapidly
than the Taylor series expansion of F (u)χ(q) given in eq.(3.14). This, in turn requires
that
lim Rn (−p) = 1 . (3.23)
n→∞
Note that the dependence on the constants Cn arising from ghost correlators in (3.7) drops
out for Rn defined in eq.(3.21). Equation (3.23) will be the key equation in the analysis
that follows.

3.3 Equations for lump solutions

In order to proceed further, we need to have an ansatz for ψ(p) and χ(p). We use the
following ansatz:
ψ(p) = K exp(−αp2⊥ /2)δ(pk ) , χ(p) = γK exp(−βp2⊥ /2)δ(pk ) , (3.24)
where K, α, β and γ are constants to be determined, and (pk , p⊥ ) denotes a decomposition
of the vector p into two orthogonal subspaces of dimensions N⊥ and Nk respectively
(N⊥ + Nk = N). In position space labelled by (xk , x⊥ ) this corresponds to a configuration
which is independent of xk and has the form of a gaussian along x⊥ . Thus a natural
interpretation of this solution is that of a solitonic (Nk − 1) brane along xk . Substituting
(3.24) into eq.(3.21), we get
γ K Rn (−p1 )g (n) (0, . . . 0) exp(−βp21⊥ /2)δ(p1k )
Z n
X  n Yn
n−1 αX
= K dN p2 . . . dN pn g (n) (p1 , . . . pn )δ( pi ) exp − p2i⊥ δ(pik ) .(3.25)
i=1 2 i=2 i=2

16
Using (3.7) in eq.(3.25), performing the integrals over pik , with i = 2, · · · n, and using the
freedom of setting p1k = 0 in g (n) (p1 , . . . pn ) and Rn (−p1 ) due to the accompanying factors
of δ(p1k ) we get
 Z n n
(α + β) 2  n−2 N⊥ N⊥
X X 
γRn (−p1⊥ ) exp − p1⊥ = K d p2⊥ . . . d pn⊥ δ( pi⊥ ) exp ak−l pk⊥ · pl⊥ ,
2 i=1 k,l=1
(3.26)
where
4 α
a0 = ln − ,
n 2
 πk 
ak = ln2 sin( ) for k 6= 0 . (3.27)
n
We now multiply both sides of the equation by eip1⊥ ·x⊥ for some N⊥ dimensional vector
x⊥ and integrate over p1⊥ . Demanding that in the n → ∞ limit Rn should approach 1,
we get
 2π N⊥ /2 x2⊥
γ exp(− )
α+β 2(α + β)
h Z Y
n n
X  n
X i
n−2
= lim K dN⊥ pk⊥ δ( pi⊥ ) exp ip1⊥ · x⊥ + ak−l pk⊥ · pl⊥ . (3.28)
n→∞
k=1 i=1 k,l=1

In order to do the integral over pi⊥ on the right hand side, we take help of the discrete
Fourier transform. We introduce new variables φl⊥ as follows:
n
1 X
φl⊥ = √ pk⊥ e−2πi(k−1)l/n , 1 ≤ l ≤ n. (3.29)
n k=1

Then n
1 X
pk⊥ = √ φl⊥ e2πi(k−1)l/n . (3.30)
n l=1
The reality condition on φl⊥ is of the form:

φ∗l⊥ = φ(n−l)⊥ for 1 ≤ l ≤ (n − 1), φ∗n⊥ = φn⊥ . (3.31)

Let us first take n to be odd. In that case we can introduce a new set of variables ξs⊥ ,
ηs⊥ through the relations:
1 n−1
φs⊥ = √ (ξs⊥ + iηs⊥ ) for 1 ≤ s ≤ . (3.32)
2 2

17
We take φn⊥ , ξs⊥ and ηs⊥ as independent variables, and express the right hand side of
eq.(3.28) in terms of these variables. This gives
 2π N⊥ /2 x2⊥
γ exp(− )
α+β 2(α + β)
 Z (n−1)/2
Y √
= lim K n−2 dN⊥ φn⊥ (dN⊥ ξs⊥ dN⊥ ηs⊥ )δ( nφn⊥ )
n→∞
s=1

i  √ X (n−1)/2 (n−1)/2
X 
2 2 2
exp √ (φn⊥ + 2 ξs⊥ ) · x⊥ − bn φn⊥ − bs (ξs⊥ + ηs⊥ ) , (3.33)
n s=1 s=1

where
n−1
X  2πks  n α n−1
X  πk   2πks 
bs = − ak cos = ln + − ln 2 sin( ) cos . (3.34)
k=0 n 4 2 k=1 n n

We can now perform the integrals over φn⊥ , ξs⊥ and ηs⊥ explicitly. This gives
 2π N⊥ /2 x2⊥ h n 1  (n−1)/2
Y π oN⊥  1 (n−1)/2
X 1 i
n−2 2
γ exp(− ) = lim K √ exp − x⊥ .
α+β 2(α + β) n→∞ n s=1 bs 2n s=1 bs
(3.35)
Using the symmetry bs = bn−s , we can rewrite this equation as
2π N⊥ /2 x2⊥ h n 1  n−1
Y π oN⊥ /2  1 2 n−1
X 1 i
γ exp(− ) = lim K n−2 exp − x⊥ .
α+β 2(α + β) n→∞ n s=1 bs 4n s=1 bs
(3.36)
For n even, we take the independent real variables to be φn⊥ , φ 2 ⊥ , ξs⊥ and ηs⊥ for
n

1 ≤ s ≤ (n − 2)/2. Repeating the analysis, we arrive at exactly the same equation (3.36).
Thus eq.(3.36) is valid irrespective of whether we take the limit n → ∞ keeping n even
or odd. This is the main equation that must be analyzed.

3.4 Solving the lump equations

We shall now try to examine if we can find solutions of (3.36) where α, β, γ and K are
finite. To this end let us define the functions:
n−1
Xπ 1 n−1
X 1
f (α, n) = ln − ln n , k(α, n) = . (3.37)
s=1 bs n s=1 bs

Comparing the two sides of equation (3.36) we find that

2
β = n→∞
lim − α ≡ C(α) , (3.38)
k(α, n)

18
 h 1 1 2π i
lim (n − 2) ln K + N⊥ f (α, n) − ln γ − N⊥ ln = 0. (3.39)
n→∞ 2 2 α+β
Equation (3.38) arises from comparing the coefficients of the gaussians in x⊥ in the two
sides of equation (3.36). These coefficients must coincide. Once the gaussians coincide,
the prefactors must coincide as well. Eqn. (3.39) arises by taking the logarithms of the
prefactors in (3.36).

1.5

0.5

alpha
1 2 3 4 5

Figure 1: The function A(α) for n = 50 (dashed) and n = 500 (continuous). Actually the error
is so small that the lines are essentially on top of each other.

Numerical analysis shows that the function k(α, n) defined through eq.(3.37) remains
finite in the n → ∞ limit. Hence C(α) is well defined. Furthermore, f (α, n) defined in
eq.(3.37) has the form
f (α, n) = A(α)n + B(α) + sn (α) , (3.40)
where sn (α) vanishes in the n → ∞ limit. In fact the function B(α) appears to vanish
identically, but we shall for the time being, not explicitly set it to zero. We show in
figure 1 a plot of A(α) calculated numerically as f (α, n + 1) − f (α, n) for n = 50 (in
dashed lines) and for n = 500 (in continuous lines). Actually the error is so small that
the lines are essentially on top of each other. In figure 2 we plot the quantity

B(n, α) ≡ 2f (α, n) − f (α, 2n) = B(α) + 2sn (α) − s2n (α) . (3.41)

With sn → 0 as n → ∞, for sufficiently large n we have that B(n, α) approaches B(α).

Thus a nonzero B(α) would show as a constant function in the plot of B(n, α). Never-
theless, the numerical evidence is that B(n, α) goes to zero accurately as n is made large.

19
 B

0.015

0.01

0.005
alpha
1 2 3 4 5
-0.005

-0.01

Figure 2: The function B(n, α) (equation (3.41)) plotted for n = 50 (dashed) and n = 500
(continuous).

1.5

0.5

alpha
1 2 3 4 5

Figure 3: The function C(α) (eqn. (3.38)) for n = 50 (dashed) and n = 500 (continuous).
Actually the error is so small that the lines are essentially on top of each other.

Indeed, as can be seen in figure 2 even for n = 50 (dashed line) B(n, α) is quite small over
the considered range. When n = 500 (continuous line) B(n, α) becomes much smaller.
We have explored this vanishing taking n ∼ 5000 and have concluded that with very high
accuracy (about one part in ten thousand)

B(α) ≃ 0 , (3.42)

20
for the range of α shown in the figure. Having concluded that B(α) = 0, the graph of
B(n, α) gives us information about the sn (α) terms. We have found that the leading
sn (α) term is of order O(1/n). With sn (α) ∼ 1/n, and for large n we have that B(n, α) ≃
3sn (α)/2 (see (3.41)). Thus figure 2 is giving us a plot of the leading sn (α) term. Note,
that there is a value ᾱ for which B(n, ᾱ) = 0. We determined numerically this value to
be ᾱ ≃ 2.772588 and believe that it corresponds exactly to ᾱ = 4 ln(2). Moreover the
dependence of sn (α) on α seems linear to high accuracy for a large range around ᾱ (see
figure). We have determined that
1 0.3333
sn (α) ≃ (α − ᾱ) + O(1/n2 ) , ᾱ = 4 ln(2) . (3.43)
2 n
The coefficient 0.3333 may really take the exact value of 1/3 but we do not know for
certain. Our investigations also suggest that for α = ᾱ the 1/n2 corrections vanish while
the 1/n3 corrections are nonzero. It is hoped that these observations may help develop
an analytical approach to this large n problem. Finally in figure 3 we show the function
C(α). This concludes our discussion of the properties of (3.40).

Substituting (3.40) into (3.39) and comparing terms linear in n as well as the constant
terms in both sides of the equation we get:9
1
ln K = − N⊥ A(α) ,
2  
1 2π
ln γ = N⊥ − ln + B(α) + 2A(α) . (3.44)
2 α+β

For a given value of α, the above, together with β = C(α) (eqn. (3.38)) determine the
functions ψ(p) and χ(p) through eq. (3.24). These functions, in turn, generate a solution
of the equations of motion of the form φ0 χ(p) + h(p) (eqn (3.20)).
As already pointed out, the natural interpretation of this solution is that of a (Nk −1)-
brane; we shall refer to this as a solitonic (Nk − 1)-brane. We would like to identify this
with D-(Nk − 1)-brane. However, since there seems to be a one parameter family of
solutions labelled by the parameter α, we have the embarassment of riches. We shall now
argue that the full solution φ0 χ(p) + h(p) is independent of the choice of α. To this end,
note that using different values of α corresponds to moving along different paths in the
complex J(p) plane and returning to the origin. Indeed, we are considering the subspace
9
Note that the existence of K and γ satisfying eq.(3.39) depends crucially on the asymptotic form
(3.40) of f (α, n). This, in turn, is a non-trivial consequence of the specific form of the n-tachyon vertex
given in eq.(3.7). For example, if g (n) had been momentum independent then we would get bs = α/2 and
f (α, n) = (n − 1) ln(2π/α) − ln n. This does not have the form (3.40).

21
of the full J(p) space characterized by the parameters α and u as follows:
2
J(p) = ue−N⊥ A(α)/2 e−αp⊥ /2 δ(pk ) . (3.45)

In the complex (u, α) plane we go around u = ub and return to u = 0 for a fixed α. For
u = 0 which is the initial and final value of u, different α correspond to same J(p), so the
different paths characterized by different values of α have the same initial and final points.
As we change α we change the path in the complex (u, α) space keeping fixed the initial
and the final points. Since we do not cross any branch point during such deformation of
the path, we should expect to have the same change in φc [p, J] for different α. As a result
the final solution φ0 χ(p) + h(p) should be independent of α, even though individually χ(p)
and h(p) depend on α. In the absence of knowledge of the h(p) term, however, we cannot
determine the explicit form of the solution.

3.5 Computation of ratios of lump tensions

We shall now compute the tension associated with this solution and compare with the
known tension of the D-(Nk − 1)-brane. For this we need to evaluate the effective action
Γ[φ] at φ = φ0 χ(p) + h(p). Since at the solution of the classical equations of motion
Γ[φ] = −W [J], the value of the action for this solution is given by:

Γ[φ0 χ(p) + h(p)] = − lim′ W [uψ(p)] . (3.46)

u→0

In taking the u → 0′ limit, we need to choose the branch in the complex u plane along
which F (u) approaches φ0 . Now from eq.(3.1) we see that on the trivial branch,
∞ Z Y
n  n
X un X
W [uψ(p)] = dN pk g (n) (p1 , . . . pn )ψ(p1 ) . . . ψ(pn )δ( pi ) . (3.47)
n=2 n! k=1 i=1

Using the form (3.7) for g (n) and (3.24) for ψ(p), and explicitly performing the integration
over the pik , we get
∞
X Z Y
n n
X n
X 
1
W [uψ(p)] = δk (0) Cn u n K n dN⊥ pk⊥ δ( pi⊥ ) exp ak−l pk⊥ · pl⊥ . (3.48)
n=2 n! k=1 i=1 k,l=1

ak ’s have been defined in eq.(3.27), and δk (0) stands for δ(pk = 0). Using the standard
finite volume regularization we can identify δk (0) with Vk /(2π)Nk , where Vk is the total
volume spanned by the coordinates xk . The integral appearing in eq.(3.48) is the same
integral that appears in eq.(3.28) with x⊥ set to zero. Performing the integral as before,
we get
∞ n  n−1
Y π oN⊥ /2
Vk X 1 n n 1
W [uψ(p)] = C n u K . (3.49)
(2π)Nk n=2 n! n s=1 bs

22
We now use eqs.(3.37), (3.40), (3.44) to write this as
∞ hN i
Vk X 1 n ⊥
W [uψ(p)] = C n u exp (B(α) + s n (α)) , (3.50)
(2π)Nk n=2 n! 2
where the coefficients sn , introduced in (3.40) approach zero as n → ∞. If we define
∞
X 1
G(u) = Cn u n , (3.51)
n=2 n!
and ∞
N X 1  
⊥
P (u) = exp B(α) Cn un eN⊥ sn (α)/2 − 1 , (3.52)
2 n=2 n!
then we can rewrite (3.50) as:
Vk h N
⊥
 i
W [uψ(p)] = exp B(α) G(u) + P (u) . (3.53)
(2π)Nk 2
Since limn→∞ sn = 0, the coefficients of the Taylor series expansion in u of P (u) grows at
a rate slower than that of G(u).
Comparison with eq.(3.14) shows that F (u) = dG(u)/du. Thus G(u) must also have
a branch point at u = ub. On the other hand, since the Taylor series expansion of P (u)
converges faster than that of G(u), P (u) has no branch point for |u| < |ub|, and may or
may not have a branch point at u = ub . We now analytically continue u around ub and
return to the origin. Suppose during this process G(u) returns to some constant G0 . It is
important to note that G0 does not depend on N⊥ , as is clear from eqn. (3.51). If P (u)
does not have a branch point at ub then it returns to zero, otherwise it returns to some
value P0 . Thus using (3.46) and (3.53) we get
Vk  N⊥ B(α)/2 
Γ[φ0 χ(p) + h(p)] = − e G 0 + P 0 . (3.54)
(2π)Nk
We shall refer to P0 as the sub-leading contribution since it comes from the function
P (u) with a milder singularity at ub than G(u). Note that P (u) and hence P0 vanishes
for N⊥ = 0. Thus if P0 does not vanish it is necessarily N⊥ dependent. We shall now
make the final assumption in our analysis, namely that the sub-leading contribution P0
vanishes.10 Using this critical assumption in eq.(3.54), we can identify the tension of the
solitonic (Nk − 1) brane to be:
1
TNk −1 = G0 e(N −Nk )B(α)/2 , (3.55)
(2π)Nk
10
This is not in contradiction with the earlier claim (previous subsection) that the sub-leading con-
tribution h(q) to the classical solution must be non-vanishing so that φ0 χ(q) + h(q) is independent of
α. Indeed, P0 receives contribution only from the sub-leading terms sn in f (α, n), whereas h(q) receives
contributions also from the sub-leading terms in k(α, n).

23
where we have used N = Nk + N⊥ . We now note that the function G(u) and hence the
constant G0 does not depend on N⊥ . Thus eq.(3.55) gives the prediction:

1 TNk −2
= eB(α)/2 . (3.56)
2π TNk −1

Using the numerical result that B(α) vanishes identically, we see that the right hand side
of eq.(3.56) becomes 1. This is in perfect agreement with the exact answer.
With our present knowledge, however, we cannot prove that P0 vanishes. Nevertheless,
the remarkable agreement described above can be taken as an evidence for the underlying
assumptions in our analysis, namely that the quadratic term in the string field theory
action expanded around the tachyon vacuum is made solely of the ghost operators, and
also that the non-leading contribution from the function P (u) vanishes at u = 0. Further
analysis of the non-leading terms sn is required to establish if P0 vanishes. In fact, even
the knowledge of Cn may be important for this analysis, in which case a specific choice of
the kinetic operator would be required.

4 Further studies and open questions

In this section we discuss important issues for which our understanding is fairly incom-
plete. We begin by exploring how much freedom there is in choosing a kinetic operator
built exclusively from the ghost sector. Then we discuss some attempts to use level expan-
sion to understand the proposed actions around the tachyon vacuum. We discuss some
difficulties with the use of the Siegel gauge, and some alternative gauges. Finally we give
some remarks on extensions to superstring theory, and on implications for the search of
closed strings states.

4.1 On the uniqueness of Q

Our discussion of kinetic operators Q that are constructed purely of ghost operators
P
identified a family of them, and thus the general form in (2.19) Q = n≥0 an Cn . More
general Q’s made of purely ghost operators might exist, but we will focus here on this
family.
The key question is whether all these operators are actually equivalent. If they are,
anyone of them could be chosen, making Q = c0 the most obvious choice. If they are not
equivalent, one must search for the correct one.
As discussed in section 2.2 and in particular in equations (2.12), field redefinitions pre-
serving the cubic structure of the theory arise from BPZ odd ghost number zero operators

24
that are derivations of the star algebra. There is a well-known family of such operators,
they are conformal transformations that leave the string midpoint fixed [75]
Kn ≡ Ln − (−1)n L−n , n ≥ 1. (4.1)
With Cn = cn + (−)n c−n we have the algebra
[Kn , Cm ] = −[(2n + m) Cm+n + (−)n (2n − m) Cm−n ] . (4.2)
For example [Kn , C0 ] = −4n Cn , and this indicates that the operator Q = C0 can be
deformed infinitesimally in every direction except along itself by a field redefinition of the
form Ψ → exp(ǫn Kn )Ψ. A change of Q along itself corresponds to a change in coupling,
and we do not expect such changes to be possible by field redefinitions.
For more general Q, even the infinitesimal deformation problem is somewhat non-
P
trivial. Consider an arbitrary Q = n≥0 an Cn and an arbitrary deformation into Q′ =
P
Q + ǫ∆Q with ∆Q = n≥0 en Cn , where only a finite number of en ’s are nonzero. We
P
ask if Q and Q′ are equivalent up to an infinitesimal scaling. With K = n dn Kn such
equivalence will hold if there exist constants dn ’s and a constant r such that
eǫK Qe−ǫK = (1 + ǫ r)Q′ → [K, Q] = rQ + ∆Q . (4.3)
If the solution involves an infinite number of non-vanishing dn ’s, one must study their
large n behaviour to decide if the infinitesimal conformal transformation associated to
K exists. While it follows from the earlier remarks that for Q = c0 all infinitesimal
deformations with e0 = 0 yield equivalent kinetic operators, we suspect that this will not
be the case for general Q. One way to see this is to recall that the identity string field
hI| is not annihilated by c0 [15]:
hI| c0 6= 0 . (4.4)
This is some sort of anomaly, since any derivation of the star product, such as c0 , would be
expected to annihilate the identity of the product. On the other hand ([15], eqn. (6.16))
 1 
hI| c0 + (c2 + c−2 ) = 0 . (4.5)
2
Since hI|Kn = 0 for all n, we cannot expect that
X X
(C0 + C2 + ǫ C0 ) ∼ exp( dn Kn ) (C0 + C2 ) exp(− d n Kn ) (4.6)
with the exponentials defining regular transformations. This is clear since the right hand
side would annihilate the identity while the left hand side would not.
It would therefore appear that up to rescaling there are at least two classes of inequiv-
alent kinetic terms, those which annihilate the identity and those which do not. In fact,
ref. [49] restricts its attention to those operators that annihilate the identity, as this string
field plays an important role in the conjectured purely cubic form of SFT.

25
4.2 On the use of level truncation
For the standard cubic open string field theory with the usual BRST operator QB , there
is overwhelming numerical evidence (but yet no proof) that the level truncation scheme
[9] provides results that converge to the correct non-perturbative answer. It is therefore of
interest to ask whether level truncation can be used to test possible choices for the BRST
operator Q around the tachyon vacuum. So far we have only obtained some partial and
somewhat problematic results for the simplest choice Q = Ã c0 . We allow for the presence
of an unknown normalization constant Ã. By a rescaling of the string field, we choose to
write Q = c0 and put an overall costant A = Ae3 in front of the string field theory action.
We first look for a translationally invariant solution of the truncated string field theory
which corresponds to the space filling D-25 brane – the original vacuum before condensa-
tion. The potential for a static configuration is
1
− S(Ψ) = V(Ψ) (4.7)
g02 2π 2
with  
1 1
V(Ψ) = A · 2π 2 hΨ, c0 Ψi + hΨ, Ψ ∗ Ψi . (4.8)
2 3
By construction, the zero of the energy corresponds to the tachyon vacuum Ψ = 0,
V(Ψ = 0) = 0 and all physical excitations are expected to have positive energy above this
ground state. We adopt the normalization conventions of [76], in which the space filling
solution ΨD25 obeys V(ΨD25 ) = 1. This in turn should allow to fix the constant A.
We work in the Siegel gauge b0 Ψ = 0. In the level (0, 0) approximation, the effect of
replacing QB by c0 is simply to change the sign of the kinetic term. From the results in [10]
√
we readily find the translationally invariant solution Ψ(0,0) = tc c1 |0i with tc = −(4 3/9)3 .
This gives V (0,0) ≈ 0.684 A. We simply quote the results that we find at higher levels:
V (2,6) ≈ 0.1849 A, V (4,8) ≈ 0.09656 A. Unlike the case of the action with the standard
BRST operator QB , here we do not find any evidence of convergence to a finite limit in
the level truncation. Rather, the energy of the space filling brane seems to be pushed
towards zero as the level is increased.
We have also looked for solitonic solutions ΨD24 corresponding to D-24 branes, in which
the string field has the shape of a lump along a direction X and approaches asymptotically
the tachyon vacuum. As in [14], in order to implement systematically the level truncation
scheme we compactify the direction X on a circle of radius R. For each zero momentum
state |Ψii that we had before, we obtain a tower of Fourier modes |Ψi,n i, n = 0, 1, 2 . . .,
of momentum p = n/R along the X direction. Level is now defined as n2 /R2 + Ni − N0 ,
where Ni is the number operator and N0 = −1 is the number eigenvalue of the zero
momentum tachyon.

26
 √
We have considered the case R = 3 and performed the computation up to level (3, 6).
As for finite A the energy of the D25-brane appears to go to zero in our calculations, we
considered an energy ratio where this fact would not affect the result. Indeed, the ratio
(N,M )
(N,M ) R V(ΨD24 )
R = (N,M )
, (4.9)
V(ΨD25 )

does not depend on the unknown constant A and is expected to converge to 1. We

find: R(1/3,2/3) ≈ 1.11056, R(4/3,8/3) ≈ 0.958355, R(7/3,14/3) ≈ 1.07661, R(3,6) ≈ 1.2141.
Although not far from the expected answer, it is not clear we are finding convergence to
the correct value.
The conclusion is that level truncation with Q = c0 and in the Siegel gauge is somewhat
problematic. This could be (i) a general problem with level expansion around the tachyon
vacuum whenever the kinetic term has been brought to a purely ghost form, (ii) a problem
with the specific choice Q = c0 as opposed to other choices in the general family Q =
P
an Cn , or, (iii) a problem with the Siegel gauge. Indeed, there are reasons to believe
that this may not be a legal choice, as we shall discuss in the next subsection.

4.3 On gauge fixing

Our analysis of the tachyon effective action in section 3 assumes that given a string field
theory with a purely ghost BRST operator Q, one can find a suitable gauge fixing which
does not involve matter operators. We believe on general grounds that this should always
be possible. However we would like to point out some complications which arise in the
application of the standard Siegel gauge to string field theories with non-standard BRST
P
operators, like the family Q = n≥0 an Cn .
We focus again on the simplest case of the string field theory with Q = c0 . Let us
consider imposing the Siegel gauge condition b0 Ψ = 0. This is always possible at the
linearized level, since if b0 Ψ 6= 0, then ΨS = Ψ − Q(b0 Ψ) = Ψ − c0 b0 Ψ satisfies b0 ΨS = 0.
Moreover, the Siegel condition fixes the gauge completely. Indeed if δΨ is pure gauge,
and in Siegel gauge, then Q δΨ = 0 and b0 δΨ = 0, leading necessarily to δΨ = 0.11
Consider now computing perturbative amplitudes in this string field theory. We shall
find that tree level Green’s functions diverge in the Siegel gauge. The only modification
of the conventional Feynman rules [41] is that the propagator is b0 instead of b0 L−1 0 .
−1 R ∞ −tL
This is however a dramatic difference. The factor L0 = 0 e 0
had the interpretation
of an integral over strips of any length. In the present case, the strip has collapsed to
zero length, and there is no integration over moduli space, but simply a ghost insertion
11
As long as L0 Ψ 6= 0, this analysis is entirely parallel to the one with QB , for which {b0 , QB } = L0 .

27
 R
b0 = dσ(bzz + bz̄ z̄ ) across the collapsed strip. Any term in the n-point amplitude requires
(n − 3) ghost insertions of the form
Z   df (n) 2 
(n) k
dz b(fk (z)) z + a.h. (4.10)
γ/2 dz
(n)
where fk (z) were defined in (3.4). Each integral across a collapsed strip has been referred
to the local coordinates z of the relevant half-strings. One of these contributions is what
is shown in the above formula. Thus γ/2 means either the right or the left half-boundary
of a canonical upper-half unit disk. In addition, the b(z) appearing in this half-string
integral has been mapped into the global uniformizing disk where all the strings are glued
together. This is necessary because all correlators are computed on the uniformizing disk.
Near z = i the integrand behaves as ∼ (1+iz)−2+4/n , which is a non-integrable singularity
for n ≥ 4. Thus in Siegel gauge, the coefficients Cn in (3.2) appear to diverge.
There is a simple cure for this problem, which consists of considering a gauge-fixing
H
condition B Ψ = 0 with B = zg(z)b(z), where g(z) ∼ (1 + iz)m near z ∼ i, and m > 1.
We also demand that g(z) contains a constant in its Laurent expansion, so that B has
some amount of b0 : this is required to have {B, c0 } =6 0 so that the gauge condition can
be imposed (at least at the linearized level). For example, with g(z) = 1 + (z 2 + 1/z 2 )/2
we have B = b0 + (b2 + b−2 )/2. The extra double zero ∼ (1 + iz)2 of the propagator near
z = i ensures now the convergence of string amplitudes. It would be interesting to explore
these ‘finite’ gauge choices in the level truncation scheme.

4.4 On superstring field theory and closed strings

As was the case for the bosonic theory we may conjecture that at the tachyon vacuum
the kinetic term of superstring field theory does not involve a matter sector operator. We
will offer some remarks that may help in constructing an appropriate action.
The algebraic structure of the open Neveu-Schwarz (NS) superstring field theory of
refs. [55, 56, 57] demands the existence of two Grassmann odd operators QB and η0
anticommuting with each other, both of which square to zero, satisfy the same BPZ
conjugation property, and are derivations of the star algebra. The operator η0 , having
picture number minus one, is the zero mode of the Grassmann odd field η(z) arising from
fermionization of the superghosts. If the BRST operator QB is replaced by c0 or any
of the Cn operators, the algebraic structure will remain in place, for this operator would
still commute with η0 and, as we have discussed before, satisfies all the other requisite
P
properties. Thus an action with Q = n an Cn would satisfy the conditions of gauge
invariance. Moreover, there would be no physical states around this vacuum. Indeed, in

28
this string field theory the linearized equations of motion are Qη0 Φ = 0 where Φ is a string
field of ghost number zero and picture number zero in the so-called large Hilbert space.
The gauge invariance δΦ = η0 Ω allows one to write Φ = ξ0 V where V is a conventional
picture number minus one field of the NS sector. In this partial gauge the equation of
motion reduces to QV = 0 which, for Q = QB is the standard BRST cohomology problem
P
in the conventional small Hilbert space. If Q = an Cn , we will have no physical states.
While it seems tempting to simply replace QB by the appropriate ghost operator to
obtain an action around the tachyon vacuum, there are some questions that suggest that
this change may not suffice to obtain the desired action. In particular, the Chan-Paton
like factors that implement the Z2 symmetry under which the GSO odd states change
sign [64, 65] must somehow change in the tachyon vacuum. Indeed, while the tachyon
potential is even with respect to the field variable which vanishes at the unstable vacuum,
it is not even with respect to the field variable which vanishes at any one of the tachyon
vacua.
P
It is perhaps worthwhile to note that the replacement QB → an Cn does not preserve
the algebraic structure of cubic superstring field theory [75]. The problem is that the
picture changing operator X(z), that must be inserted at the string midpoint in the
definition of the star product, while BRST invariant, is not Cn invariant. This spoils
the derivation property which is necessary for gauge invariance. This cubic string field
theory, in any case, has long been known to be problematic [77], and does not appear to
give a sensible tachyon potential either [78]. Interestingly, the double picture changing
insertion used in the cubic superstring field theory action advocated in refs.[69, 70, 72]
does commute with the Cn operators, and it therefore appears that gauge invariant actions
without cohomology could be constructed.
We can now ask if a similar replacement of the BRST operator with another having
trivial cohomology can be done in closed string field theory [80]. The answer is no. The
problem here is that the interactions beyond the cubic one involve the integration of cor-
relators over finite pieces of the moduli spaces of Riemann surfaces. Key to the algebraic
structure of the theory is the fact that the BRST operator acts as a total derivative on
moduli space. This arises because forms on moduli space involve the antighost field b(z)
and {QB , b(z)} = T (z), where the energy momentum tensor T generates deformations of
correlators. Such property will not hold upon replacing QB with an object constructed
purely from ghosts.
A final question is whether closed strings arise from the kind of open string field theory
we have been considering here, namely one with a kinetic opertor having no cohomology.
We have no definite answer. Recall that in cubic open string field theory closed string
poles definitely arise in string loop amplitudes [81]. A subset of open string propagators

29
in a string diagram, represented by a strip of total length T ∈ [0, ∞], whose middle line
defines a nontrivial closed curve, will give rise to closed string poles via integration over
the neighborhood of T = 0. In the kind of actions we have considered here, the propagator
collapses to an overlap with some antighost insertion. It does not seem possible to obtain
closed string poles from an amplitude that just includes the contribution from the point
T = 0. On the other hand, there are well known complications with the precise Batalin-
Vilkovisky quantization of conventional cubic open string field theory, and they precisely
arise from the T → 0 limit of a one-point one-loop amplitude [82, 83]. The whole issue
surrounding closed strings in open string field theory after tachyon condensation is subtle
enough [84, 85, 86, 87, 88] that further work seems necessary to attain definite conclusions.

4.5 Concluding remarks

The open string field theories studied in this paper manifestly implement the absence of
conventional open string dynamics on the vacuum of the tachyon. They naturally imple-
ment gauge invariance and are rather reminiscent of the p-adic string theory, where after
a simple invertible field redefinition the kinetic operator is just a constant. In addition,
they are in many ways strikingly simple, much more so than the conventional cubic open
string field theory in the background of a D-brane. For example, while we were able here
to compute the exact momentum dependence of the off-shell n-tachyon amplitudes, such
computation appears to be prohibitively complicated in conventional cubic string field
theory. If the normalization of the n-tachyon amplitudes can be computed, perhaps along
the lines of comments in sect. 4.3, we may then be able to find the explicit solutions
representing D-branes, to obtain the precise form of the kinetic term Q, and to settle the
outstanding assumptions of our analysis.

Acknowledgements: We would like to thank D. Ghoshal, D. Jatkar, N. Moeller, P.

Mukhopadhyay, S. Panda, and W. Taylor for useful discussions. The work of L.R. was
supported in part by Princeton University “Dicke Fellowship” and by NSF grant 9802484.
The work of B.Z. was supported in part by DOE contract #DE-FC02-94ER40818.

References
[1] A. Sen, Descent relations among bosonic D-branes, Int. J. Mod. Phys. A14, 4061
(1999) [hep-th/9902105].

[2] A. Sen, Non-BPS states and branes in string theory, hep-th/9904207.

30
 [3] A. Recknagel and V. Schomerus, Boundary deformation theory and moduli spaces of
D-branes, Nucl. Phys. B545, 233 (1999) [hep-th/9811237].

[4] C.G. Callan, I.R. Klebanov, A.W. Ludwig and J.M. Maldacena, Exact solution of a
boundary conformal field theory, Nucl. Phys. B422, 417 (1994) [hep-th/9402113].

[5] J. Polchinski and L. Thorlacius, Free fermion representation of a boundary conformal

field theory, Phys. Rev. D50, 622 (1994) [hep-th/9404008].

[6] P. Fendley, H. Saleur and N.P. Warner, Exact solution of a massless scalar field with
a relevant boundary interaction, Nucl. Phys. B430, 577 (1994) [hep-th/9406125].

[7] J. Harvey, D. Kutasov and E.J. Martinec, On the relevance of tachyons, [hep-
th/0003101].

[8] S. Dasgupta and T. Dasgupta, Renormalization group analysis of tachyon condensa-

tion, hep-th/0010247.

[9] V.A. Kostelecky and S. Samuel, The Static Tachyon Potential in the Open Bosonic
String Theory, Phys. Lett. B207 (1988) 169.

[10] A. Sen and B. Zwiebach, Tachyon Condensation in String Field Theory, JHEP 0003,
002 (2000) [hep-th/9912249].

[11] N. Moeller and W. Taylor, Level truncation and the tachyon in open bosonic string
field theory, Nucl. Phys. B583, 105 (2000) [hep-th/0002237].

[12] J.A. Harvey and P. Kraus, D-Branes as unstable lumps in bosonic open string field
theory, JHEP 0004, 012 (2000) [hep-th/0002117].

[13] R. de Mello Koch, A. Jevicki, M. Mihailescu and R. Tatar, Lumps and p-branes in
open string field theory, Phys. Lett. B482, 249 (2000) [hep-th/0003031].

[14] N. Moeller, A. Sen and B. Zwiebach, D-branes as tachyon lumps in string field theory,
hep-th/0005036.

[15] L. Rastelli and B. Zwiebach, Tachyon potentials, star products and universality, hep-
th/0006240.

[16] A. Sen and B. Zwiebach, Large marginal deformations in string field theory, JHEP
0010, 009 (2000) [hep-th/0007153].

31
[17] W. Taylor, Mass generation from tachyon condensation for vector fields on D-branes,
JHEP 0008, 038 (2000) [hep-th/0008033].

[18] I. Ellwood and W. Taylor, private communication and work in progress.

[19] R. de Mello Koch and J.P. Rodrigues, Lumps in level truncated open string field
theory, hep-th/0008053.

[20] N. Moeller, Codimension two lump solutions in string field theory and tachyonic
theories, hep-th/0008101.

[21] A. Kostelecky and R. Potting, Analytical construction of a nonperturbative vacuum

for the open bosonic string, hep-th/0008252.

[22] H. Hata and S. Shinohara, BRST invariance of the non-perturbative vacuum in

bosonic open string field theory, JHEP 0009, 035 (2000) [hep-th/0009105].

[23] B. Zwiebach, Trimming the Tachyon String Field with SU(1,1), hep-th/0010190.

[24] M. Schnabl, Constraints on the tachyon condensate from anomalous symmetries,

hep-th/0011238.

[25] R. Gopakumar, S. Minwalla and A. Strominger, Noncommutative solitons, JHEP

0005, 020 (2000) [hep-th/0003160].

[26] K. Dasgupta, S. Mukhi and G. Rajesh, Noncommutative tachyons, JHEP 0006, 022
(2000) [hep-th/0005006].

[27] J. A. Harvey, P. Kraus, F. Larsen and E. J. Martinec, D-branes and strings as non-
commutative solitons, hep-th/0005031.

[28] E. Witten, Noncommutative tachyons and string field theory, hep-th/0006071.

[29] J. A. Harvey, P. Kraus and F. Larsen, Exact noncommutative solitons, hep-

th/0010060.

[30] G. Mandal and S. Rey, A note on D-branes of odd codimensions from noncommutative
tachyons, Phys. Lett. B495, 193 (2000) [hep-th/0008214].

[31] A.A. Gerasimov and S.L. Shatashvili, On exact tachyon potential in open string field
theory, hep-th/0009103.

[32] D. Kutasov, M. Marino and G. Moore, Some exact results on tachyon condensation
in string field theory, hep-th/0009148.

32
[33] D. Ghoshal and A. Sen, Normalisation of the background independent open string
field theory action, hep-th/0009191.

[34] J. R. David, Tachyon condensation using the disc partition function, hep-th/0012089.

[35] D. Ghoshal and A. Sen, Tachyon condensation and brane descent relations in p-adic
string theory, Nucl. Phys. B584, 300 (2000) [hep-th/0003278].

[36] B. Zwiebach, A solvable toy model for tachyon condensation in string field theory,
JHEP 0009, 028 (2000) [hep-th/0008227].

[37] J.A. Minahan and B. Zwiebach, Field theory models for tachyon and gauge field string
dynamics, JHEP 0009, 029 (2000) [hep-th/0008231].

[38] J. A. Minahan and B. Zwiebach, Effective tachyon dynamics in superstring theory,

hep-th/0009246.

[39] J. A. Minahan and B. Zwiebach, Gauge fields and fermions in tachyon effective field
theories, hep-th/0011226.

[40] G. Arutyunov, S. Frolov, S. Theisen and A. A. Tseytlin, Tachyon condensation and

universality of DBI action, hep-th/0012080.

[41] E. Witten, Noncommutative Geometry And String Field Theory, Nucl. Phys. B268,
253 (1986).

[42] E. Witten, On background independent open string field theory, Phys. Rev. D46,
5467 (1992) [hep-th/9208027].

[43] E. Witten, Some computations in background independent off-shell string theory,

Phys. Rev. D47, 3405 (1993) [hep-th/9210065].

[44] K. Li and E. Witten, Role of short distance behavior in off-shell open string field
theory, Phys. Rev. D48, 853 (1993) [hep-th/9303067].

[45] S.L. Shatashvili, Comment on the background independent open string theory, Phys.
Lett. B311, 83 (1993) [hep-th/9303143].

[46] S.L. Shatashvili, On the problems with background independence in string theory,
hep-th/9311177.

[47] A. A. Tseytlin, Sigma model approach to string theory effective actions with tachyons,
hep-th/0011033.

33
[48] E. Witten, unpublished.

[49] G. T. Horowitz, J. Morrow-Jones, S. P. Martin and R. P. Woodard, New Exact

Solutions For The Purely Cubic Bosonic String Field Theory, Phys. Rev. Lett. 60,
261 (1988).

[50] G. T. Horowitz, J. Lykken, R. Rohm and A. Strominger, A Purely Cubic Action For
String Field Theory, Phys. Rev. Lett. 57, 283 (1986).

[51] H. Hata, K. Itoh, T. Kugo, H. Kunitomo and K. Ogawa, “Pregeometrical String

Field Theory: Creation Of Space-Time And Motion,” Phys. Lett. B175, 138 (1986).

[52] R. Gopakumar, S. Minwalla and A. Strominger, Symmetry restoration and tachyon

condensation in open string theory, hep-th/0007226.

[53] N. Seiberg, A note on background independence in noncommutative gauge theories,

matrix model and tachyon condensation, hep-th/0008013.

[54] A. Sen, Some issues in non-commutative tachyon condensation, hep-th/0009038.

[55] N. Berkovits, Super-Poincare Invariant Superstring Field Theory, Nucl. Phys. B450
(1995) 90, [hep-th/9503099].

[56] N. Berkovits and C.T. Echevarria, Four-Point Amplitude from Open Superstring Field
Theory, [hep-th/9912120].

[57] N. Berkovits, A New Approach to Superstring Field Theory, proceedings to the

32nd International Symposium Ahrenshoop on the Theory of Elementary Particles,
Fortschritte der Physik (Progress of Physics) 48 (2000) 31, [hep-th/9912121].

[58] A. Sen, Stable nonBPS bound states of BPS D-branes, JHEP 08, 010 (1998) hep-
th/9805019.

[59] A. Sen, Tachyon condensation on the brane anti-brane system, JHEP 08, 012 (1998)
hep-th/9805170.

[60] A. Sen, SO(32) spinors of type I and other solitons on brane - anti-brane pair, JHEP
09, 023 (1998) hep-th/9808141.

[61] E. Witten, D-branes and K theory, JHEP 12, 019 (1998) hep-th/9810188.

[62] A. Sen, BPS D branes on nonsupersymmetric cycles, JHEP 12, 021 (1998) hep-
th/9812031.

34
[63] P. Horava, Type IIA D-branes, K theory, and matrix theory, hep-th/9812135.

[64] N. Berkovits, The Tachyon Potential in Open Neveu-Schwarz String Field Theory,
[hep-th/0001084].

[65] N. Berkovits, A. Sen and B. Zwiebach, Tachyon condensation in superstring field

theory, hep-th/0002211.

[66] P. De Smet and J. Raeymaekers, Level four approximation to the tachyon potential
in superstring field theory, hep-th/0003220.

[67] A. Iqbal and A. Naqvi, Tachyon condensation on a non-BPS D-brane, hep-

th/0004015.

[68] L. Brekke, P. G. Freund, M. Olson and E. Witten, Nonarchimedean String Dynamics,

Nucl. Phys. B302, 365 (1988).

[69] I. Y. Arefeva, P. B. Medvedev and A. P. Zubarev, Background Formalism For Su-

perstring Field Theory, Phys. Lett. B240, 356 (1990).

[70] I. Y. Arefeva, P. B. Medvedev and A. P. Zubarev, New Representation For String

Field Solves The Consistence Problem For Open Superstring Field, Nucl. Phys. B341,
464 (1990).

[71] C. R. Preitschopf, C. B. Thorn and S. A. Yost, Superstring Field Theory, Nucl. Phys.
B337, 363 (1990).

[72] I. Y. Aref’eva, D. M. Belov, A. S. Koshelev and P. B. Medvedev, Tachyon conden-

sation in cubic superstring field theory, hep-th/0011117.

[73] A. Sen and B. Zwiebach, A proof of local background independence of classical closed
string field theory, Nucl. Phys. B414, 649 (1994) [hep-th/9307088].

[74] T. Kawano and T. Takahashi, Open string field theory on noncommutative space,
hep-th/9912274.

[75] E. Witten, Interacting Field Theory Of Open Superstrings, Nucl. Phys. B276, 291
(1986).

[76] A. Sen, Universality of the tachyon potential, JHEP 9912, 027 (1999) [hep-
th/9911116].

35
[77] C. Wendt, Scattering amplitudes and contact interactions in Witten’s superstring field
theory, Nucl. Phys. B 314 209 (1989).
J. Greensite and F.R. Klinkhamer, Superstring Amplitudes and Contact Interactions,
Nucl. Phys. B304 (1988) 108.

[78] P. De Smet and J. Raeymaekers, The tachyon potential in Witten’s superstring field
theory, JHEP 0008, 020 (2000) [hep-th/0004112].

[79] S. Giddings, E. Martinec and E. Witten, Modular invariance in string field theory,
Phys. Lett. B226 362 (1986).
B. Zwiebach, A proof that Witten’s open string theory gives a single cover of moduli
space, Commun. Math. Phys. 142 193 (1991).

[80] B. Zwiebach, Closed string field theory: Quantum action and the BV master equation.
Nucl.Phys. B390(1993) 33-152, hep-th/9206084.

[81] D. Z. Freedman, S. B. Giddings, J. A. Shapiro and C. B. Thorn, “The Nonplanar One

Loop Amplitude In Witten’s String Field Theory,” Nucl. Phys. B298, 253 (1988).

[82] C. B. Thorn, String field theory, Phys. Rep. 174 (1989) 1;

A. R. Bogojevic, BRST Invariance of the measure in string field theory, Phys. Lett.
B198 (1987) 479;
M. Maeno, Canonical quantization of Witten’s string field theory using midpoint light
cone time, Phys. Rev. D43 (1991) 4006.

[83] B. Zwiebach, Oriented open-closed string theory revisited, Annals Phys. 267, 193
(1998) [hep-th/9705241].

[84] M. Kleban, A. Lawrence and S. Shenker, Closed strings from nothing, hep-
th/0012081.

[85] P. Yi, Membranes from five-branes and fundamental strings from Dp branes, Nucl.
Phys. B550, 214 (1999) [hep-th/9901159].

[86] O. Bergman, K. Hori and P. Yi, Confinement on the brane, Nucl. Phys. B580, 289
(2000) [hep-th/0002223].

[87] G. Gibbons, K. Hori and P. Yi, String fluid from unstable D-branes, hep-th/0009061.

[88] A. Sen, Fundamental strings in open string theory at the tachyonic vacuum, hep-
th/0010240.

36